Share Email Print
cover

Proceedings Paper • new

Speckle reduction in optical coherence tomography using two-step iteration method (Conference Presentation)
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Optical coherence tomography (OCT) provides high resolution and cross-sectional images of biological tissue and is widely used for diagnosis of ocular diseases. However, OCT images suffer from speckle noise, which typically considered as multiplicative noise in nature, reducing the image resolution and contrast. In this study, we propose a two-step iteration (TSI) method to suppress those noises. We first utilize augmented Lagrange method to recover a low-rank OCT image and remove additive Gaussian noise, and then employ the simple and efficient split Bregman method to solve the Total-Variation Denoising model. We validated such proposed method using images of swine, rabbit and human retina. Results demonstrate that our TSI method outperforms the other popular methods in achieving higher peak signal-to-noise ratio (PSNR) and structure similarity (SSIM) while preserving important structural details, such as tiny capillaries and thin layers in retinal OCT images. In addition, the results of our TSI method show clearer boundaries and maintains high image contrast, which facilitates better image interpretations and analyses.

Paper Details

Date Published: 19 April 2017
PDF: 1 pages
Proc. SPIE 10057, Multimodal Biomedical Imaging XII, 1005702 (19 April 2017); doi: 10.1117/12.2250832
Show Author Affiliations
Xianghong Wang, Nanyang Technological Univ. (Singapore)
Xinyu Liu, Nanyang Technological Univ. (Singapore)
Nanshuo Wang, Nanyang Technological Univ. (Singapore)
Xiaojun Yu, Nanyang Technological Univ. (Singapore)
En Bo, Nanyang Technological Univ. (Singapore)
Si Chen, Nanyang Technological Univ. (Singapore)
Linbo Liu, Nanyang Technological Univ. (Singapore)


Published in SPIE Proceedings Vol. 10057:
Multimodal Biomedical Imaging XII
Fred S. Azar; Xavier Intes, Editor(s)

© SPIE. Terms of Use
Back to Top